Add missing documentation to solver.h
Change-Id: I86e7c4f1f6cc1e15d5eb2cf23e73c32d94d458c1
diff --git a/include/ceres/solver.h b/include/ceres/solver.h
index 88358bc..b8be006 100644
--- a/include/ceres/solver.h
+++ b/include/ceres/solver.h
@@ -160,8 +160,23 @@
// the inverse of the Hessian matrix. The rank of the
// approximation determines (linearly) the space and time
// complexity of using the approximation. Higher the rank, the
- // better is the quality of the approximation. For more details,
- // please see:
+ // better is the quality of the approximation. The increase in
+ // quality is however is bounded for a number of reasons.
+ //
+ // 1. The method only uses secant information and not actual
+ // derivatives.
+ //
+ // 2. The Hessian approximation is constrained to be positive
+ // definite.
+ //
+ // So increasing this rank to a large number will cost time and
+ // space complexity without the corresponding increase in solution
+ // quality. There are no hard and fast rules for choosing the
+ // maximum rank. The best choice usually requires some problem
+ // specific experimentation.
+ //
+ // For more theoretical and implementation details of the LBFGS
+ // method, please see:
//
// Nocedal, J. (1980). "Updating Quasi-Newton Matrices with
// Limited Storage". Mathematics of Computation 35 (151): 773–782.
